Dead Labour: The Human Ghosts inside AI
By combining AI with ‘digital remains,’ we get the curious phenomenon of ‘dead labour’. Image by Ian Mikraz on Unsplash+.
The dead live on inside most of the technologies we use.
For most of human history, the traces left behind by the dead were fragile and finite: letters in boxes, names etched into stone, photographs curling at the edges in attic heat. What survived depended largely on accident, geography, and the devotion of the living.
But digital technologies have transformed both the scale and persistence of human traces. Today, the dead remain present in messages, voice notes, videos, social media posts, emails, cloud archives, playlists, search histories, and endless streams of data generated almost continuously throughout life. Technology is no longer simply a tool for remembering the dead. Increasingly, it is where the dead gather.
As a cyberpsychologist, I’ve spent years thinking and writing about digital afterlives: what happens when human identity, memory, and relationship become entangled with technological systems that preserve, sort, monetise, and recirculate our traces long after bodily death. More recently, I’ve found myself thinking about dead labour, also known as ‘spectral labour’ or even ‘spectral capitalism.’
Traditionally associated with Marxist critiques of industrial capitalism, ‘dead labour’ referred to the accumulated labour of past workers embedded into machines, infrastructure, and capital itself. But in the age of artificial intelligence, the phrase has begun to feel strangely literal. Today’s AI systems are built not only upon the ongoing labour of the living, but upon immense archives of preserved human expression: the writing, artwork, conversations, emotional disclosures, recordings, preferences, and behavioural traces of countless people, many of whom are no longer alive.
Book a Keynote on Digital Afterlife: Grief, Technology, and AI
As you're reading, keep in mind that I deliver keynotes, facilitate workshops, and consult on digital afterlives explore my speaking offerings or get in touch.
More of my writing on the psychology and ethics of digital afterlives:
What Is ‘Dead Labour’?
Karl Marx used the phrase ‘dead labour’ to describe the way past human work becomes embedded into machines, systems, and capital. A factory machine, for example, did not emerge from nowhere. It contained within it the accumulated labour of the people who designed it, built it, maintained it, and refined it over time. Over centuries, through what they built and left behind, we can continue to extract value from the labour of the dead.
Artificial intelligence intensifies this idea, and we’re now in a curious era of new-style dead labour for the digital age, sometimes termed ‘spectral labour’ or even ‘spectral capitalism.’
Today’s AI systems are built from immense archives of human expression and behaviour. They are trained on human writing, human art, human voices, human photographs, human conversations, human emotional disclosures, human relationships, human histories. Increasingly, they are trained on the traces we leave behind simply by working, communicating, and existing online.
These traces often persist independently of us. Sometimes, indeed often, they persist after death for an indefinite period of time. We constantly underestimate just how much we will leave behind digitally, and usually have no concept of what can be done with it after we are dead, what profits can continue to be be extracted from us.
The humans whose digital traces become raw material for these systems possess little meaningful agency over how those traces are used. Our words, images, preferences, and interactions are continuously datafied, commodified, and recirculated through systems we neither control nor fully understand. The digital self becomes less a stable possession than a resource to be extracted, and that extraction continues long after we cease breathing.
And, unlike earlier industrial forms of dead labour, today’s systems are built not only from human productivity, but from human intimacy. Our grief, vulnerabilities, relationships, desires, and identities increasingly become part of the substrate from which intelligent systems are constructed. Long after we are dead, our digital remains will continue to refine the rapidly emergent ‘attachment economy,’ and used to train more human-seeming ‘empathic AI’ and hyperrealistic chatbots.
The Dead Already Live in Our Systems
We’ve always used technology to reach towards the dead. The telegraph inspired séances. The emergence of photography quickly led to the business of spirit photography. Edison imagined a phonograph sensitive enough to capture the voices of the war dead. The technologies changed, but the human instinct remained consistent: we always use the tools available to us to maintain bonds with those we have lost.
Now, something has fundamentally shifted. Technology is no longer simply a medium through which we remember the dead. Increasingly, it is where the dead live on. Wherever the living have gathered online for work or socialising, pleasure or for profit, the dead will also be present.
The modern dead persist in messages, photographs, videos, emails, social media posts, voice notes, search histories, playlists, cloud archives, location histories, and behavioural data generated across years or decades of digital life. The traces of a person no longer disappear gradually into the fragility of paper, memory, or physical decay. They remain active within systems designed explicitly to preserve, sort, analyse, recirculate, and recombinate them.
These archived traces of the dead continue participating in economic systems long after bodily death. Human expression persists as data. Data persists as infrastructure. Infrastructure persists as value.
The dead continue to generate engagement. Their images remain searchable. Their writing remains scrapeable. Their conversations remain analysable. Their voices remain reproducible. Increasingly, their accumulated traces may also continue training AI systems, becoming part of the statistical substrate from which new synthetic language, images, personalities, and interactions emerge.
In this sense, the dead are not merely remembered by digital systems. They are operational within them.
This profoundly complicates older ideas about mortality, privacy, consent, and identity. Historically, death eventually imposed limits. The body decayed. Memory faded. Artefacts disappeared or became inaccessible. But digital systems are structured against forgetting. Platform architectures are designed for retention, extraction, and recirculation. They are, in many respects, anti-oblivion machines.
At the same time, these systems encourage us to think about ourselves less as embodied, relational beings and more as collections of retrievable data points. We become profiles, archives, metrics, behavioural patterns, datasets. The phrase ‘data subject’ sounds administrative and harmless, but its implications are profound. Linguistically and psychologically, it subtly reframes personhood itself. We are not encountered primarily as living creatures moving through the world, but as informational resources.
The same logic extends beyond the living. The digital dead are data subjects too, and this gives us a greater sense of permission to make free with their remains, to use them as we will.
Humans have always sought forms of symbolic survival. What’s historically unprecedented is that these traces now persist inside extractive commercial systems that continuously convert identity, memory, grief, and human expression into economic assets. The afterlife has become part of the infrastructure of everyday digital life.
Invisible Humans Beneath ‘Artificial’ Intelligence
Artificial intelligence conceals the living humans inside it too, as Kate Crawford explains in The Atlas of AI.
The dominant cultural imagery surrounding AI is curiously disembodied. We imagine sleek interfaces, glowing servers, elegant automation, machine ‘intelligence’ operating frictionlessly at enormous scale. The language surrounding AI systems reinforces this illusion constantly: seamless, autonomous, efficient, self-improving. The technology is often presented almost as though it has emerged independently of human bodies and human environments altogether.
But this is a fantasy.
Beneath the polished surfaces of AI systems lies an immense hidden infrastructure of human labour. Data labellers categorise images and text so systems can learn patterns. Content moderators spend their working lives exposed to disturbing and traumatic material in order to keep platforms usable for everyone else. Click workers perform repetitive microtasks for minimal pay. Warehouse workers, lithium miners, engineers, delivery drivers, and factory labourers sustain the physical infrastructures upon which supposedly ‘virtual’ technologies depend.
The cleaner and more frictionless the system appears, the easier it becomes to psychologically distance ourselves from the humans sustaining it. As we use the systems they suffer so much to support, perhaps we hasten their deaths. We don’t know. We can’t seen them.
When labour becomes invisible, people become easier to conceptualise as functions rather than subjects. The content moderator becomes part of the platform architecture rather than a human nervous system repeatedly exposed to horror. The data labeler becomes merely ‘training data infrastructure.’ The person disappears into the system they sustain.
AI systems encourage us to focus attention on the apparent intelligence of the machine rather than on the vast accumulation of human life, labour, and vulnerability embedded within it. The fantasy of machine autonomy depends upon a collective forgetting of the humans underneath the machine: both the living and the dead.
What Happens When Data Outlives the Body
When it comes to the actual digital dead, I don’t think the central psychological question here is whether AI using ‘dead labour’ or ‘spectral labour’ is good or bad. We have always stood on the shoulders of the dead, continued to extract value from them in various ways. In a knowledge or information economy, the labour of the past is concretised in the digital remains we leave behind.
Perhaps this is comforting, in some sense. We like to think that our life’s work will continue to matter. Symbolic or actual immortality has always tended to be a tantalising idea. We wish to be remembered, to extend our influence, to predict we’ll have ongoing power or importance when we’re no longer here. Perhaps it helps offset death anxiety, the terror of nonbeing.
But what happens psychologically when we increasingly encounter one another as archives or echoes of once-living people, rather than embodied beings? When preserved fragments of human expression continue circulating independently of the humans who produced them?
Historically, death imposed a kind of rupture. The dead persisted symbolically, emotionally, spiritually perhaps, but not operationally, not to the extent they persist today. The traces of the dead continue moving and speaking and influencing through algorithmic systems. Data outlives embodiment. Digital remains become hypervisible, while the people themselves fade from view.
I don’t know exactly where this leads us. Perhaps future generations will experience the dead as a meaningful continuation of human presence and memory. Perhaps these systems will deepen connection with generations gone before. Perhaps we will find ways of making the actual humans behind AI-delivered information more visible, of giving them more credit. I rather hope so.
For more on the emergent phenomenon of ‘dead labour,’ watch this space. There are countless examples of it across every sector, sometimes quite literal: dead professors, resurrected actors, posthumously preaching politicians. When it comes to labour, the line between the living and the dead is increasingly thin.
FAQs About Dead Labour
What is 'dead labour'?
Karl Marx used 'dead labour' to describe the past human work absorbed into machines, infrastructure, and capital. A factory machine carries the accumulated effort of everyone who designed, built, and maintained it. The living keep drawing value from work the dead already did.
What are 'spectral labour' and 'spectral capitalism'?
They update the idea of dead labour for the AI age. Today's systems train on vast archives of human writing, art, voices, and behaviour, much of it produced by people who have died. Their preserved traces become raw material for new technologies, so the work of the dead carries on long after the body stops.
How does AI use the data of people who have died?
Messages, images, recordings, work outputs, and all other behavioural traces persist in all the platforms we use, which are built for retention rather than forgetting. That material can be scraped, analysed, and fed into models that generate new text, images, and synthetic personalities. The dead remain searchable, reproducible, and active inside systems they never agreed to join.
What happens to your digital traces after you die?
Most people leave behind far more than they imagine, across messages, photos, cloud archives, playlists, and search and location histories. Few have any say over what becomes of it. Companies can keep extracting value from those traces for an indefinite period, with little consent from the person who produced them or the people who survive them.
Who are the hidden human workers behind AI?
As Kate Crawford argues in The Atlas of AI, the frictionless surface of AI conceals an enormous human workforce: data labellers who categorise images and text, content moderators exposed to traumatic material, click workers paid little for repetitive microtasks, and the miners, engineers, and warehouse staff who sustain the physical infrastructure. The cleaner the system looks, the easier it becomes to forget them. We forget both the living and the dead that underpin the AI we use, and they have something in common: invisibility and powerlessness.
Why does treating people as 'data subjects' matter psychologically?
The phrase sounds administrative, but it reframes personhood. Encountering one another as profiles, archives, and behavioural patterns rather than as living, embodied people changes how we relate, and it extends to the dead, whose remains we then feel freer to use as we wish.
Is using the data of the dead to train AI wrong?
Humans have always built on the work of those who came before. What has shifted is that the traces of the dead now persist inside commercial systems that convert our personhood into assets or ore to be mined; transform them; extract from them; and utilise them independently of the consent or knowledge of people who created them, whether those people be living or dead.